Best web scraping APIs, 2026Every main kind of web scraping API.Scraping APIs, crawl-and-extract APIs for AI, search APIs and browser APIs, compared on what they return, how they charge and how you connect.
An editorial comparison of what each vendor publishes, checked 25 Sept 2026. Monocrawl is ours and listed first. How we compared
What are you building?
Clean Markdown from pages and whole sites, for an LLM app or RAG pipelineAn option that returns Markdown and maps or crawls a whole site for you. They differ most in how they charge.
An AI agent that reads web pages as it worksAn option with its own MCP server and a tool that reads pages, so the agent can call it directly.
Pages from sites that block scrapers, fetched at volumeA scraping API with proxies, JavaScript rendering and retries built in. Most charge more for premium proxies or rendering.
Product, article or listing fields as JSON, from a schema or a promptAn option with AI extraction: you describe the fields and get JSON back.
Clicks, typing and scrolling in a hosted browser, for pages a plain fetch can’t readA browser API that Playwright or Puppeteer connects to, or session calls that run page JavaScript.
14options compared
12with an MCP server
14free to start
60sources, checked 25 Sept 2026
Four routes
Crawl-and-extract APIs return clean Markdown or JSON from a page or a whole site. Scraping APIs fetch pages past blocks at volume, and search APIs for AI find pages before reading them. Browser APIs give you a hosted browser to drive.
Monocrawl, Firecrawl, Spider, Jina Reader and Apify
ScrapingBee, ScraperAPI, Zyte, Bright Data and others
Tavily and Exa
Google SearchMonocrawl, Bright Data, Firecrawl, Scrapfly and others
What they return
What each option lists on its own pages, and how you connect. Structured extraction is JSON from a schema or prompt; a batch job takes many URLs in one call; a browser session is a hosted browser you control.
| Scraping features | Connect | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Option | Page scrape | JavaScript rendering | Site crawl | Site map | Structured extraction | Markdown output | Batch jobs | Browser sessions | REST API | MCP server | Sources |
| MonocrawlOur product | Our API | ||||||||||
| ApifyCrawler and scraper actors | – | 12345 | |||||||||
| FirecrawlCrawl and extract API | 678910 | ||||||||||
| Jina ReaderURL-to-Markdown API | – | – | – | – | 1112 | ||||||
| SpiderCrawl and scrape API | 1314151617 | ||||||||||
| Bright DataUnlocker, browser and crawl APIs | 1819202122 | ||||||||||
| DecodoWeb scraping API | – | – | – | 2324252627 | |||||||
| Scrape.doWeb scraping API | – | – | – | – | 2829303132 | ||||||
| ScraperAPIWeb scraping API | – | 3334353637 | |||||||||
| ScrapflyScraping and browser APIs | 3839404142 | ||||||||||
| ScrapingBeeWeb scraping API | – | 4344454647 | |||||||||
| Zyte APIWeb scraping API | – | – | – | 4849505152 | |||||||
| ExaSearch and contents API | – | – | – | 535455 | |||||||
| TavilySearch API for AI agents | – | – | – | 5657585960 | |||||||
Pricing
A request, a result, a record and a unit are not the same size, and allowances reset on different schedules. Price your own workload with each vendor.
One key and one response format across 70+ platforms, with monitors for alerts.
Monthly plans
$0.20 per compute unit on the free and Starter plans
Apify’s tests put Website Content Crawler at $0.50 to $5 per 1,000 pages with a headless browser. Starter plan $19 a month.
Worth knowingWebsite Content Crawler, maintained by Apify, switches between plain HTTP and a headless browser to suit each page.
$5 of usage a month
$19 a month for 5,000 credits, billed monthly
A basic scrape, crawl or map costs 1 credit a page; JSON extraction adds 4. $16 a month billed yearly.
Worth knowingA scrape that returns nothing is not charged, but a page that answers with an error such as 403 still costs 1 credit.
1,000 credits a month
$50 for 1 billion tokens, shared across Jina’s APIs
Reader charges by the length of the content it returns. $500 buys 11 billion tokens.
Worth knowingPrepend r.jina.ai to a URL to get its main content back as Markdown; s.jina.ai searches the web.
Keyless use at 20 requests a minute; 10M tokens for non-commercial use
$1 per GB of page content, plus $0.0001 per minute of compute
Pay as you go with no subscription, and the balance never expires.
Worth knowingBills bandwidth and compute rather than pages, so JavaScript rendering carries no fixed multiplier.
Keyless page scrapes, rate-limited
$1.50 per 1,000 Web Unlocker requests, pay as you go
$499 a month includes 383,000 requests. Its Browser API costs $8 per GB, and the Crawl API $1.50 per 1,000 records.
Worth knowingIts Crawl API returns a whole site as Markdown, text, HTML or JSON from one root URL.
5,000 requests a month
$0.75 per 1,000 requests with JavaScript rendering, on its $19 plan
$0.50 per 1,000 without it; premium proxies with JavaScript cost $1.50. Prices exclude VAT.
Worth knowingEach request’s price depends on two choices: standard or premium proxies, and JavaScript rendering on or off.
Free plan: 2,000 requests a month, 1,000 with JavaScript
$29 a month for 250,000 credits
A request costs 1 credit, or 5 with JavaScript rendering. Only successful requests are charged.
Worth knowingIts Async API runs in a separate pool, worth 30% of your plan’s concurrency, alongside regular requests.
1,000 credits a month
$49 a month for 100,000 API credits, billed monthly
A plain request costs 1 credit and one with JavaScript rendering 10. $44.10 a month billed yearly.
Worth knowingIts AI Parser learns a site’s layout from up to three example pages, then returns JSON for any similar page.
7-day trial with 5,000 credits; free plan of 1,000 credits
$30 a month for 200,000 credits
A basic request costs 1 credit and one with JavaScript rendering 5. Failed requests are free.
Worth knowingOne credit balance covers its scraping, cloud browser, screenshot, extraction and crawler APIs.
1,000 credits, no card
$19 a month for 75,000 API credits
A request costs 5 credits with JavaScript rendering, its default, or 1 without; AI extraction adds 5. Prices exclude VAT.
Worth knowingIts auto mode tries the cheapest proxy and rendering setup first and charges only the one that succeeds.
1,000 free credits
$0.13 per 1,000 HTTP responses from the easiest sites, pay as you go
Up to $1.27 for the hardest. Browser-rendered responses cost $1.01 to $16.08 per 1,000.
Worth knowingIts automatic extraction returns product, article and job posting fields without selectors, and can add attributes you define.
$5 of credit, for 30 days
$1 per 1,000 pages of contents, per content type
AI page summaries add $1 per 1,000 pages. Search is $7 per 1,000 requests.
Worth knowingPage contents come from its own cache first, with a live crawl as the fallback for pages it has not stored.
$10 of credits a month
$30 a month for 4,000 credits
Five page extractions cost 1 credit, as does a basic search. Pay as you go is $0.008 a credit.
Worth knowingYou are not charged for URL extractions or map requests that fail.
1,000 credits a month
Try a call
The same request in cURL, TypeScript or Python, and the recorded map of www.iana.org with a limit of 10: 10 links back, starting with iana.org/about, iana.org/domains and iana.org/domains/root. No crawler to run and no sitemap to parse.
GET /v1/web/mapcurl "https://www.monocrawl.com/v1/web/map?url=https%3A%2F%2Fwww.iana.org&limit=10&sitemap=skip" \
-H "x-api-key: mn_your_key_here"Links on iana.org
iana.org/about, iana.org/domains, iana.org/domains/root
Selected public fields from the recorded response. Live values will differ.
Page as Markdown1 creditSite map, up to 500 URLs3 creditsCrawl, up to 50 pages5 creditsBatch, up to 25 URLs3 creditsMetadata and JSON-LD1 creditPDF or document text2 creditsBrowser session, up to 60 seconds5 creditsBrowser agent task10 credits2.1sMedian response in our benchmark. 22 September 2026 catalogue benchmark. Median of recorded endpoint timings across 636 operations with a successful HTTP response. Includes network time; individual requests vary.
Questions
Most let you start free. Every month, Monocrawl gives 1,000 credits, Bright Data 5,000 requests and Decodo 2,000 requests, or 1,000 with JavaScript; Firecrawl, Tavily and Scrape.do give 1,000 credits each, Exa $10 of credits and Apify $5 of usage. ScrapingBee and Scrapfly give 1,000 credits once, ScraperAPI a 7-day trial with 5,000 credits, and Zyte $5 of credit. Jina Reader and Spider also read pages without an API key, at a limited rate.
List the site’s URLs, then read each page as Markdown, or run one crawl that does both. With Monocrawl, web/map finds up to 500 URLs from a site’s sitemap, web/scrape returns a page as Markdown and web/crawl collects up to 50 pages with a Markdown excerpt of each. Firecrawl, Spider, Tavily and Apify’s Website Content Crawler crawl whole sites into Markdown, Jina Reader converts one URL when you prepend r.jina.ai, and scraping APIs such as ScrapingBee, ScraperAPI, Scrape.do and Decodo offer a Markdown output option.
Monocrawl, Apify, Bright Data, Decodo, Exa, Firecrawl, Jina Reader, ScraperAPI, Scrapfly, ScrapingBee, Spider and Tavily publish their own MCP servers with tools that read web pages, so assistants such as Claude or Cursor can call them directly. Zyte publishes a guide to building your own MCP server on its API. Through Monocrawl’s MCP server an agent can scrape, map and extract pages; crawl and batch jobs, the browser agent and browser sessions start over its REST API.
Most do, and most charge more for it. ScrapingBee renders by default at 5 credits a request, against 1 without; ScraperAPI charges 10 credits and Scrape.do and Scrapfly 5, against 1 for a plain request. Decodo charges $0.75 per 1,000 results against $0.50, and Zyte’s browser-rendered responses start at $1.01 per 1,000 against $0.13. Firecrawl’s 1-credit scrape handles JavaScript-rendered pages, and Spider bills rendering as compute. Monocrawl’s web/scrape reads static HTML by default; its browser sessions and browser agent run JavaScript in a hosted browser.
Options with crawl, site map and Markdown endpoints of their own include Monocrawl, Spider, Tavily and Apify, whose Website Content Crawler and Sitemap Extractor are maintained by Apify. Jina Reader converts one URL at a time to Markdown. Scraping APIs such as ScrapingBee, ScraperAPI and Scrapfly pair proxy-backed page fetches with Markdown output and AI extraction, and Bright Data, Scrapfly and Spider also sell hosted browsers. Firecrawl itself starts free with 1,000 credits a month.
Units differ: a credit, a request, a result, a page, a token, a gigabyte and a compute unit are not the same size. Many vendors multiply the cost of a request for JavaScript rendering, premium proxies or AI extraction, some also bill pages that return errors, and allowances reset on different schedules. Price your own workload, such as a number of pages a month on the sites you need, with each vendor.
Method
What’s includedWidely used web scraping, crawling and browser APIs, and search APIs for AI that also read pages. Each comes up across web scraping API, web crawler API and Firecrawl alternative searches and comparison lists, documents its API publicly and publishes its pricing or offers a trial.
What we comparedWhat each vendor publishes about scraping, JavaScript rendering, crawling, extraction and browser features, pricing, free allowance and access, read on its own pages. “Best” here means worth comparing, not a ranking from tests: we have not run their APIs.
About MonocrawlMonocrawl is our product and is listed first. Its details come from our live plans and catalogue.
1,000 free credits every month. No card required.
Start free